Scholarlink

<p>Errors rarely occur in isolation, but rather within behavioral sequences and are known to shape subsequent behavior. However, understanding the full time course of the impact of errors has been challenging given data limitations. Here, to gain key insights, two groups of precisely matched participants (<i>n</i> &gt; 35,000) were compared; those who made versus did not make an error in a short object discrimination task. Linear mixed-effects models revealed error-induced effects both preceding and following errors. First, the error group showed pre-error speeding up to eight trials before the error occurred. Second, post-error slowing occurred and rapidly diminished over trials. Finally, leave-one-out cross-validation models using individual trial behavior predicted an upcoming error eight trials before the error occurred. This project highlights the systematic nature of behavioral changes around errors and demonstrates the potential for early error prediction and intervention using only simple metrics and models.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting errors before they occur using trial-by-trial dynamics of pre-error and post-error behavior

  • 期刊论文

  • 发表日期：

    2026年5月29日

查看全文
  • Sarah B. Malykke,
  • Audrey Siqi-Liu,
  • Dwight J. Kravitz,
  • Stephen R. Mitroff

摘要

Errors rarely occur in isolation, but rather within behavioral sequences and are known to shape subsequent behavior. However, understanding the full time course of the impact of errors has been challenging given data limitations. Here, to gain key insights, two groups of precisely matched participants (n > 35,000) were compared; those who made versus did not make an error in a short object discrimination task. Linear mixed-effects models revealed error-induced effects both preceding and following errors. First, the error group showed pre-error speeding up to eight trials before the error occurred. Second, post-error slowing occurred and rapidly diminished over trials. Finally, leave-one-out cross-validation models using individual trial behavior predicted an upcoming error eight trials before the error occurred. This project highlights the systematic nature of behavioral changes around errors and demonstrates the potential for early error prediction and intervention using only simple metrics and models.